View Full Version : CRAY Jaguar Super Computer surpases IBM's RoadRunner as the fastest cluster ...
toupeiro
November 17th, 2009, 02:07 AM
... in the world. CRAY is using 6-core AMD powered opterons. IBM's uses a hybrid of their own CELL processor and Opteron. I loved the part of the article where it took a stab at the marketing ploys already taking place due to this:
AMD: 4 out of the 5 top 500 compute clusters use AMD.
Intel: 402 of the top 500 super compute clusters run Intel.
I don't think there were any pure Nehalem architecture clusters big enough to compete with the top 5. I'm hoping to see how their numbers hold respective to how many nodes were used.
Discuss! Try and keep it clean. :)
http://www.engadget.com/2009/11/16/cray-jaguar-leaps-past-ibm-roadrunner-as-worlds-faster-supercom/
Ian dewhurst
November 17th, 2009, 02:14 AM
I'm still staggered by the fact petaflops were to be achieved around 2015 (or so it was claimed) and its done way ahead of schedule.
Its definitely one-upmanship in respect of the AMD vs Intel lawsuit a nice piece of PR if I do say so myself.
Grant A.
November 17th, 2009, 02:33 AM
Whether or not I care depends on this question:
Will it be cheaply available for my computer in the next coming years?
MaxIBoy
November 17th, 2009, 03:18 AM
Cray still exists?
Anyway, we can see that even for supercomputers, AMD is competitive. (And that Linux is still the champ.)
Exodist
November 17th, 2009, 03:32 AM
Cray is expensive. Its cheaper to build a Beowulf cluster. But then again software must be written for a cluster to make real use of it.
toupeiro
November 17th, 2009, 03:49 AM
Cray still exists?
Anyway, we can see that even for supercomputers, AMD is competitive. (And that Linux is still the champ.)
Cray doesn't just exist. They've completely rebounded and redefined themselves.
However, a cray system for your home, probably not unless you're just that well off. ;)
Frak
November 17th, 2009, 07:21 AM
Cray doesn't just exist. They've completely rebounded and redefined themselves.
However, a cray system for your home, probably not unless you're just that well off. ;)
We have a CX1 where I work.
toupeiro
November 17th, 2009, 07:31 AM
We have a CX1 where I work.
Very nice! Do you get to work on it?
At my site we have a very small 18 node compute cluster, infiniband interconnected. Our corporate offices has a compute cluster with node numbers in the thousands last time I saw it. They're built internally, but not sure what kind of performance numbers they produce in comparison with the ones produced by Jaguar or Roadrunner. I know we have some tesla systems as well,but I don't get to admin those machines. :(
Frak
November 17th, 2009, 04:17 PM
Very nice! Do you get to work on it?
I wish. It has a schedule. I work on it, sorta, but not directly on it.
8:00 AM - 10:00 AM : CAD Renders.
10:00 AM - 12:00 PM : Animator Renders.
12:00 PM - 1:00 PM : Lunch, so it goes through the queue to work on not yet processed jobs still in the pool.
1:00 PM - 3:00 PM : Distributed compilation.
3:00 - 5:00 PM : First come, first serve. If nobody is requesting use, it goes through the pool.
5:00 PM - 8:00 AM : Stop taking jobs, complete the remaining jobs in the pool, and shutdown. Start back up at 7:55 AM.
We need another one now that everybody relies on it so much.
The Real Dave
November 17th, 2009, 10:02 PM
Who says you can't have a supercomputer at home?
For $10,000 NVIDIA say you can! :D NVIDIA Personal Supercomputing (http://www.nvidia.com/object/personal_supercomputing.html)
Powered by vBulletin® Version 4.2.2 Copyright © 2024 vBulletin Solutions, Inc. All rights reserved.